Survey instrument

Following a brief introduction, the survey instrument consists of three main parts: CE choice sets, images of nature items and attitudes items, as well as socio-demographic questions. Instrument development was aided by extensive qualitative interviews (n= 13).

Pre-test (n=50) was conducted in Beijing in autumn 2005. With improved instrument and survey administration, a larger scale precursor study was administered in Beijing and Chengdu (n=213 for CE, and n=101 for images of nature items). Furthermore, an additional qualitative interview (n=9) were conducted. Based on the comprehensive analysis of the qualitative and quantitative data, as well as interviewer and respondent feedbacks, the final version of survey instrument was designed.

3.3.1 Images of nature items

Our images of nature items stem from two sources: (i) the dissertation of Krömker (2004, 2005) and (ii) original items operationalising a “harmony” dimension. Krömker’s items focus on eight aspects of the human-nature-relation:

• Purpose: plants and animals exist primarily for human use.

• Robustness: Nature is not that fragile that it has to be protected by humans. It can best help itself.

• Reciprocity²: Whatever humans take from nature they must give something back in order to keep the balance of the universe.

• Nature needs & deserves to be protected: Nature should be protected because it enriches our lives.

• Spiritual: Nature is spiritual or sacred in itself.

• Threatening: Many processes of nature are dangerous to humans.

• Perverse/Tolerant: To some degree humans can modify nature. Nature gets out of control only if a specific threshold is crossed.

• Limitation: Earth is like a spaceship with only limited space and resources.

Based on an intensive literature analysis (see section 2.2), four aspects were singled out in order to design a set of harmony items that would supplement the above items to account specifically for traditional Chinese images of the human-nature relation:

• Wholeness: Nature is a whole formed by different living beings (animals and plants). Humans are a part of this unit.

• Equality: Human and nature, including animals and plants, have an equal right to exist.

• Respect: Nature has its own law. Humans should respect it and follow it in order to live in harmony.

• Proper proportion: When humans interfere with nature, they should follow the law of proper proportion.

2 Krömker terms this aspect “respect”. In fact, the item points rather towards reciprocity as the central idea, however. In order to differentiate this aspect from the “respect for nature” idea that is part of the Chinese harmony concept, we use the term “reciprocity” here.

Table 1. Examples of images of nature items

• Human and nature, including animals and plants have the equal right to exist.

(Equality and harmony)^§

• Humans must follow the law of nature in order to live in harmony. (Respect and fragility)

• Ignoring the law of the nature will eventually bring the disaster to humans. (Protection and prevention)

• Nature always recovers (by itself), no matter what humans do. (Robustness and non-spirituality)

• Nature is sensitive to any kind of interference. Even small interference ca lead to big and irreversible damage. (Spirituality and fragility)

§:Phrase in parenthesis identifies the images of nature dimension to which the item was assigned by factor analysis (see results section).

Starting from the original English and German language versions, the first author translated twenty seven images of nature items developed by Krömker (2004) into Chinese. Twelve harmony items were designed. In all, 39 items were tested in the precursor study (n=101) of which 25 were kept in the main study. We used a closed answer format with a 5-point Likert scale (1: totally disagree to 5: totally agree). For examples of images of nature items, see Table 1.

3.3.2 Attribute design

The first step of designing CE choice sets was the identification of attributes that characterized southwestern China destinations. To optimize this step, the first author conducted 22 qualitative in-depth interviews with Chinese middle class respondents who were interested in travelling to southwestern China in spring and autumn 2005. From lists of factors that tourists said they paid most attention to when making their destination decisions, the five following attributes were constructed (Table 2). The sustainable tourism services attribute was the only attribute not spontaneously mentioned by interviewees but included due to research interests.

Table 2. Destinations attributes/levels used in choice experiment

1. No car roads, only trails; no hotels or hostels, tourists need to bring their own food; (none level)

2. Difficult motor access; several simple hostels or hotels; (difficult level)

3. Old roads; some common hotels and restaurants; (limited level) 4. National roads and highways; many hotels from two stars to four or

five stars, and many restaurants; (advanced level)

1

1. Sense of solitude and tranquillity; no tourism infrastructure; no other visitors; (virgin land level)

2. At times, some basic tourism infrastructure (restaurants, stores, toilets) visible; meet other people every hour; (basic level)

3. Have big scale tourism infrastructure (cable car, Karaoke), always be aware of "this is a tourism place"; meet other people every minute;

(dispersed level)

4. Many kinds of tourism infrastructures everywhere; many "artificial"

attractions; very crowded; (packed level)

1

2. Some traditional old buildings mixed with many modern buildings; no original living culture; (endangered level)

3. Several traditional local buildings mixed with many modern houses;

one temple without usage any more; modernized local customs and culture; (modernized level)

4. Well preserved traditional local buildings; temples with monks playing roles in local people’s life; and well preserved original customs and culture; (original level)

1 2

3

4

Natural attractions^$

1. Ordinary landscapes; no precious or attractive species; (none level) 2. Magnificent awe-inspiring natural landscapes; no precious or

attractive species; ( landscape level)

3. Ordinary landscapes, many precious species and very abundant species-rich biodiversity (Panda, Golden Monkey); ( species level) 4. Magnificent awe-inspiring natural landscapes, many precious species

and very abundant species-rich biodiversity (Panda, Golden Monkey); (both level)

1. Freely visit every place; no visitor restriction; no resource saving or recycling; no community involvement; (none level)

2. Small conservation program only for core attraction area; litter/trash cans; community based business-restaurants, hostels and small hotels, local specialty stores and horse riding service; (limited level)

3. Having buffer zone; scientific conservation program for core area and buffer zone; green buses, bio energy use and waste water treatment;

local community participates in decision-making in local

development, and local business involvement (see above); (extensive level)

#base level (opt out option) was coded as 0. ^$the natural attractions level 2 and 3 are coded “2” because there is no

“natural” order of the two levels. ^§the attribute tarnished nature experience shows up in the actual choice cards (Figure 2) under the more neutral description possibility of experiencing nature.

For each attribute, three to five levels were constructed based on representative conditions of nature and landscape-based tourism attraction sites in southwestern China. The levels of the cost attribute are based on published information on trip costs ranging from trips for modest backpackers, self-organized travellers and commercially organized group-trips to

‘luxury’ trips including renting vehicles with drivers.

Some published CE studies include similar destination attributes. For example, travel convenience is included by Hearne & Salinas (2002) and Huybers (2003). An attribute comparable to our ‘tarnished’ nature experience is used by Apostolakis & Jaffry (2005), Hanley et al. (2002), and Huybers (2003). Kelly et al. (2007) deals with sustainable tourism services, while Hanley et al. (2002) or Naidoo and Adamowicz (2005) employ attributes on landscape and scenic quality.

3.3.3 CE choice sets

From the attributes and attribute levels, 3,840 (4⁴*3*5) different combinations of destination characteristics can be generated. An orthogonalisation procedure was used to generate a main effects design leading to 24 pair-wise comparisons of destination options, i.e. choice sets. The choice sets were randomly blocked into three groups. Each respondent was asked to choose the preferred trip from one of two described destinations (Cards A and B), and an opt out/buy nothing option (Figure 2).

Figure 2. Example of a choice set (English language version³)

3.4 Administration of the survey

Before the formal interview started, respondents were asked if they like travelling, where they live, and about their approximate income. If respondent income was lower than 1,500 RMB per month (~150 €), the interviewers would ask two more interactive questions and finished the interview. Next, the CE attributes were explained. With a set of sample cards, the choice procedure was practiced. Directly before the choice exercise, respondents received a token gift as the appreciation of their participation. Images of nature questions were asked after the choice excise. In a final section of the interview, further socio-demographic questions were asked. Overall, a total of 4,928 choices were observed from 616 respondents (see also Figure 3).

3 The Chinese version and English version choice sets with visual aids are available at: http://www.uni-goettingen.de/de/sh/47518.html.

3.5 Statistical and econometric analysis

3.5.1 Factor analysis and correlation analysis

Factor analyses and correlation analysis were conducted with SPSS 15.0. A varimax rotated factor analysis with binary squared Euclidean distance and Ward-linkage was employed to generate five empirical dimensions of images of nature (Kroemker, 2004). Based on factor analysis results, all items i of each dimension d with a factor loading l above 0.4 were used to calculate an individual dimension score V of the respective dimension for each respondent n:

∑

^∗

=

i

Q l n d V( , )

with Q: Likert score of i

For all dimensions, a descriptive label was chosen that reflects contents and wording of the underlying items. While these dimension scores are used for econometric analysis, we also report mean raw scores for each dimension to facilitate a more intuitive understanding with reference to the original 1-5 likert scale (Table 3).

To avoid bias non-normal score distributions, nonparametric standard procedures (bivariate Spearman correlation) were used for correlation analysis. For econometric analysis, residency and gender were dummy coded (Figure 3a and 3b). Education and self-perceived social status were coded as 1-4 and 1-5 (Figure 3d and 3e). In the survey, respondents were asked to indicate the right ranges they belong to regarding their age, monthly income and annual travel expenditures. In analysis, age, monthly income and annual travel expenditures (Figure 3c, 3f and 3g) used means of each category for calculation.

3.5.2 Nested Logit (NL) model estimation

For the analysis of the CE data, a set of Nested Logit models (NL) was calculated with NLOGIT 3.0. Preliminary analyses indicated the risk of violations of the independence from irrelevant alternatives (IIA) condition necessary for the application of (the simpler)

conditional logit analysis. Because NL does not rely on the IIA assumption, an eligible NL tree structure was identified, and the corresponding models estimated. The inclusive value was set to 1.0 for the degenerated branch, and the model initiated with starting values obtained from a non-nested NL model (Hensher et al. 2005: 536). All scale parameters were normalized at the lowest level (RU1).

All models include an alternative specific constant (ASC) coded 1 for the generic choices A and B, and 0 for the “opt out/buy nothing” option. The ASC expresses a fundamental propensity to make (or not to make) a trip to southwestern China beyond the information given during the CE.

In the tables presenting the NL models, pseudo-R² values (constant only), Log likelihood function (LL) and inclusive value (IV) are reported as diagnostic statistics. Pseudo-R² values in reference to a constant only model are much more conservative than the R² value of ordinary least squares (OLS), for example, values between 0.07-0.08 correspond to R² values of 0.22 to 0.24 value in an OLS equivalent (Hensher et al., 2005: 338). IV statistics are significantly different from 1 highlighting the insufficiency of ordinary multinomial logit models. All NL models are overall highly significant (p-value of Chi²-Test < 0.001).

3.5.3 Interaction terms

To test for the influences of images of nature dimensions on preferences, interaction terms between attributes and dimensions were generated. In a first step, these interaction terms were included one-by-one as single terms into a NL base model.

With NL procedures, we estimated an additive utility function of the form

U = b1*attribute1 + … + b6*attribute6 + bi*interaction termi + basc*ASC with

b1…b6: attribute coefficients including the cost attribute b6;

bi: coefficient of interaction termi;

basc: coefficient of the Alternative Specific Constant.

Because the coefficients of the interaction terms bi are estimated with individually varying dimension scores, these coefficients represent some of the preference heterogeneity within the sample.

3.5.4 Parsimonious model and willingness-to-pay (WTP) estimation

In a second analysis step, we combined promising interaction terms into a single NL model.

The parsimonious models of Table 7 were generated by stepwise exclusion non-significant interaction terms using the conventional significance threshold of p≤0.1. The estimation of statistically significant attribute and interaction coefficients allows for the calculation of welfare measures. The maximum willingness-to-pay for a 1 level change of trip attribute Si

(“marginal” WTP) equals the ratio of the respective coefficient bi (bi=attribute) and the negative value of coefficient of the cost attribute b6.

6

)

( b

S b

mWTP _i =− ⁱ

If the WTP calculations include interaction terms, sample mean of the dimension scores of the images of nature dimensions must be accounted for (see Table 7).

4 Results

4.1 Respondent socio-demographics

Figure 3 summarizes the socio-demographic characteristics of the sample. There are 307 respondents of Chengdu (49.8%) and 309 (50.2%) of Beijing. A gender ratio of 53:47 (male:

female) was achieved. Respondents less than 39.9 years old comprise 61% of the sample. Age groups (Figure 3c) reflect age distribution among Chinese middle class sub-districts, although large efforts were put on recruiting respondents older than 40 years. With differences between

Beijing and Chengdu, mean monthly income is 3,537 RMB (~354 €), which is about twice of the average monthly income in Beijing and four time of average monthly income in Chengdu (CNBS, 2006a). Average annual travel expenditures are ~3,500 RMB (~350 €). The amount is nearly five times higher than urban resident average tourism expenditures of 737 RMB (~74 €) in 2005 (CNTA, 2006). Half (50%) of respondents hold a formal educational degree equivalent to a bachelor degree or above. This is much higher than the Chinese average of 5.8% (CNBS, 2006b). Around 90% of the respondents regard themselves at least as middle class.

f

§: “unsure” (n=32) is substituted by the estimated values from a co-linearity diagnostics linear regression. n=616.

Figure 3. Respondent socio-demographics overview

4.2 Images of nature dimensions

Factor analysis generated six dimensions from images of nature items. Based on reliability considerations, we selected the first five dimensions (for details, see Table 3). The equality and harmony dimension explains 22% of factor analysis variance. The respect and protection value dimension explains 9.6%, 5.8% is explained by the consequence and personal connection dimension, the robustness and non-spirituality dimension accounts for 4.9% while the spirituality and fragility explains 4.5%. A total of 46.8% of variance is explained by the dimensions.

The five dimensions can be grouped into two categories that either focus on a harmonious human-nature relation or domination of nature for human purposes. With the exception of the robustness and non-spirituality dimension, all other dimensions belong to the harmony-oriented dimensions.

The dimension mean raw score (5-point Liker scale) shows that average respondents have a high level agreement with the values expressed by the harmonious dimensions (4.3, 4.3, 4.4

and 3.7). For the domination of nature dimension robustness and non-spirituality, the mean raw score is only 2.3.

Table 3. Images of nature dimensions and items

Images of nature In the grand design of world, humans have the same value with other living beings. 0.734

Human and nature, including animals and plants have the equal right to exist. 0.706 Human should protect nature because it has a right of existence in itself in the same way that all and everything living does.

0.699 Humans belong to nature the same way as animals and plants do. 0.569 As the supreme beings on earth, human should not tarnish nature. 0.507 When human interfere with nature, they should follow the law of proper portion. 0.74 Humans should protect nature because it provides recreation and quietness. 0.543 Humans must follow the law of nature in order to live in harmony. 0.511 Humans should protect nature because it enriches our lives by its wonderful magnificence. 0.471 Nature is sensitive to any kind of interference. Even small interference can lead to big and

irreversible damage.

0.462 Ignoring the law of the nature will eventually bring the disaster to humans. 0.734 Humans should protect nature because it is useful and provides a lot of advantages for us. 0.701

I feel threatened by the ongoing destruction of nature. 0.53

The earth is like a spaceship with only limited room and resources. 0.482

Nature is sacred because it is created by God. 0.59

Nature has its own right of existence; therefore it is not allowed to destroy nature anywhere for human needs.

0.497 Nature is sensitive to any kind of interference. Even small interference can lead to big and

irreversible damage.

0.453 The earth is like a spaceship with only limited room and resources. 0.429 Humans have the right to use natural resources of any kind they want to. 0.677

Nature is important, but neither has a soul nor is sacred. 0.588

Plants and animals do exist primarily for human use. 0.581

Nature always recovers (by itself), no matter what humans do. 0.573 Not humans can protect nature; only God has the power to do so. 0.486 Robustness and

4.3 Correlation of images of nature dimensions with socio-demographics

The equality and harmony dimension is positively correlated with education but negatively correlated with gender (Table 4). The respect and protection value dimension is positively correlated with age and self-perceived social status but negatively correlated with gender.

Higher educated respondents and Chengdu respondents agree more strongly with the consequence and personal connection dimension. The robustness and non-spirituality dimension has a positive correlation with age and negative correlation with education and monthly income. Female and Chengdu respondents agreed more strongly with the values expressed by the spirituality and fragility dimension.

In sum, respondents agreeing above average with the harmonious human-nature relation dimensions are more frequently from Chengdu, more often female and tend to have a higher education. Age and self-perceived social status are the only surveyed socio-demographic variables that are exclusively correlated with the respect and protection value dimension, and not shared by other harmonious dimensions. On the other hand, supporter of the human domination dimension are older, lower educated and receive less income than the average respondent.

Table 4. Correlations between images of nature dimensions and socio-demographics

Images of nature dimensions

Residency Gender Age Education Monthly Income

Annual travel expenditures

Self-perceived social status

Equality and harmony (harmonious) -0.09* 0.081*

Respect and protection value (harmonious) -0.127** 0.178** 0.089*

Consequence and personal connection

(harmonious) -0.146** 0.103*

Spirituality and fragility (harmonious) -0.082* -0.202**

Robustness and non-spirituality

(domination) 0.156** -0.172** -0.092*

Socio-demographics

**significance at 0.01 level (two-tailed); ^*significance at 0.05 level (two-tailed). In residency, Beijing is coded as 1 and Chengdu is coded as 0 in residency. In gender, male is coded as 1 and female is coded as 0.

4.4 Destination choice base model

Table 5 presents the influences of six destination attributes on respondent preferences for destination choice. All six attributes have significant influences on destination choice in southwestern China nature-based destinations. Convenience, cultural and natural attractions have highly significant and positive influences regarding destination choice. The coefficients of the tarnished nature experiences and the sustainable tourism services are significant and negative. Cost is highly significant, and displays a negative sign of the utility coefficient, as to be expected from fundamental micro-economic theory. The negative sign of the ASC coefficient indicates that actual respondent preferences are overestimated if calculated from the attribute levels alone.

Table 5. Nested logit model result for destination attributes

Attribute Coefficient P

Convenience 0.1067*** 0.0000

Tarnished nature experience -0.0652** 0.0031

Sustainable tourism services -0.0623* 0.0309

Cultural attractions 0.1884*** 0.0000

Natural attractions 0.4242*** 0.0000

Cost [1,000RMB] -0.1958*** 0.0000

[Non-status quo ASC] -0.2482* 0.0346

Log likelihood function -4933.26

Pseudo-R² (constant only)^§ 0.071

Inclusive value (IV) 0.842

Observations (choices) 4928

***: significance at 0.001 level; ^**: significance at 0.01 level; ^*: significance at 0.05 level. The ASC in brackets as it is a NL model predictor of destination choice but not a destination attribute. ^§Pseudo R² values in reference to a constant only model-values between 0.07-0.08 correspond to R² values of 0.22 to 0.24 value in for the linear model equivalent (Hensher et al. 2005:338); IV statistic is significantly different from 1; n=616.

4.5 Images of nature dimensions influences on destination choice

Table 6 displays the results from interaction analysis with the images of nature dimensions.

The respect and protection value dimension positively influences preferences for the convenience attribute (p=0.002), and negatively influences the overall value of the offered trips as compared to the opt out/buy nothing options (negative interaction with the ASC;

The respect and protection value dimension positively influences preferences for the convenience attribute (p=0.002), and negatively influences the overall value of the offered trips as compared to the opt out/buy nothing options (negative interaction with the ASC;

Im Dokument Cultural and attitudinal influences on destination choice (Seite 171-0)